Uncovering the Role of T Cells in Cancer With Spatial Biology

The effectiveness of cancer immunotherapy is hindered by mechanisms that allow tumor immune escape and subsequent tumor progression. T cells are essential to the tumor microenvironment, but a deeper understanding of the mechanisms which prevent T cells from entering tumors is crucial to develop new therapies and improve existing ones.1 This application focus explores how high-plex spatial profiling with the MACSima™ Platform can be applied to better understand the role of T cells in cancer immune surveillance and how cancers evade the immune system. Using spatial biology to uncover γδ T cell mechanisms Gamma delta (γδ) T cells are a subset of T cells that are essential effectors in tumor immunosurveillance. They bridge features of both adaptive and innate immunity, have antigen-presenting capacity and exhibit cytotoxic and regulatory functions independently of the major histocompatibility complex (MHC). However, their function and resulting clinical significance has been difficult to elucidate. Notably, it is known that γδ T cells have both antitumor function and protumor function, and their accumulation in tumor tissues has been identified as the best predictor for positive prognosis across 39 malignancies.2 There are also currently several avenues of research attempting to use γδ T cells in immunotherapies, as their lack of MHC restriction means that they do not induce adverse immune responses (i.e., graft vs host disease) when administered. However, the biological mechanisms underlying γδ T cell functions must be fully understood in order to exploit them for therapeutic approaches. MACSima Imaging Cyclic Staining (MICS) technology (Figure 1) was used in a proof-ofprinciple study to investigate γδ T cells in their native tissue microenvironment.3 Figure 1. An overview of a MACSima multiomics workflow using MACSima Imaging Cyclic Staining (MICS) technology. Credit: Miltenyi Biotec Uncovering the Role of T Cells in Cancer With Spatial Biology2 High-plex in situ proteomics imaging using the MACSima Platform The MACSima Platform automates cyclic staining to simultaneously acquire data from over 200 markers on a single tissue slide, allowing a deeper look into subphenotypes in the tissue. Any traditional antibodies used for immunohistochemistry or immunofluorescence staining or any antibody that works on fresh, frozen or FFPE tissue can be used. A serial section is taken and H&E stained. Pathologists identify regions of interest (ROIs), such as the tumor center, invasive front and normal tissue. These are then used to set the respective ROIs on the MACSima Platform (Figure 2). Subsequent stitching and alignment of all the images and removal of background noise produces high-plex images which can be analyzed with MACS® iQ View Spatial Biology analysis software to bring scientific context to the results. This is particularly useful for the visualization of the changes in cell phenotypes and the resulting interactions that enable tumors to infiltrate healthy tissue (Figure 3). This proof of principle study identified 28 γδ T cell neighborhoods from 6 benign colon samples and 22 colorectal cancer samples. The study was performed on fresh frozen tissues that were PFA-fixed on 5 μm thick tissue slides. For the assessment of the markers, high-plex imaging with the MACSima Platform was performed with 56 markers per tissue (Figure 4). Several immune cell populations were selected for the markers in the panel, as well as tissue-forming cells and cancer-associated epitopes. These markers were either selected from the broad pretested antibody portfolio covered by Miltenyi Biotec or from other vendors. In the resulting image of a fresh, frozen human colon sample, an immune infiltrating compartment can be seen, along with a muscular layer and crypts (circular structures surrounding the tertiary immune infiltration spot). Most importantly, γδ T cells were also stained, allowing their study in a high-dimensional context. Subsequently, a workflow was established to identify cell phenotypes, annotate cell clusters and bring more context into the images. A cell segmentation was performed, followed by a transformation of greyscale values and principal coefficient analysis. Finally, a Leiden clustering was performed. The resulting cell type annotation illustrated cell subset clusters, including the immune cells of interest. The findings Figure 2. Sample preparation for MACSima workflow. Credit: Kilian Wistuba-Hamprecht, Manfred Claassen Figure 3. High-plex images of human colon tissue (healthy compared to tumor invasive front) taken using the MACSima Platform.3 from the computational analysis were used to quantify the neighborhoods of the δ1 γδ cells and δ2 γδ cells (Figure 5). This showed differences in interactions between different immune cells and the γδ T cell subtypes. For example, there was a clear higher frequency of myeloid cells interaction with δ2 cells compared to δ1 cells. Furthermore, the phenotypes of colorectal cancer (CRC) and hepatocellular carcinoma (HCC) infiltrating δ1 and δ2 cells were identified, and the cell-type composition within individual neighborhoods was determined. Single-cell sequencing was used to validate the findings from the MACSima Platform, particularly the phenotypic signatures of γδ T cells at the single cell level. Deciphering the role of MAIT cells in liver cancer with multiplex tissue imaging While the previous approaches were sufficient for single-cell analysis, a more general computational approach was required to determine the broader composition and phenotypes of the Tcell neighborhoods. In a collaboration with partners at the National Institutes of Health (NIH), multiplex tissue imaging was used to understand the role of mucosal-associated invariant T (MAIT) cells in liver cancer and patient survival. MAIT cells detect intracellular metabolites as a method of identification of abnormal cells by the immune system. Previous research into the role of MAIT cells in liver cancer, specifically HCC, has shown contradictory results. While some studies showed that high levels of MAIT cells in the liver correlated with a better chance of survival, other studies showed the opposite. Therefore, the hypothesis of this study was that the tissue microenvironment (the neighborhoods of these MAIT cells) may determine their role (either beneficial or non-beneficial) in liver cancer.4 For the multiplex imaging performed in this study, 37 different protein markers were measured simultaneously on the same tissue. To understand the results, a machine learning model was used to determine differences in cell type composition of the cells surrounding MAITs in healthy tissue and tumor tissue. The neighborhoods around MAIT cells in healthy and tumor tissue were classified and the frequency of each neighboring cell type was quantified, allowing for comparison of enriched or depleted cell types between the two tissue types (Figure 6). However, this analysis can only be as accurate as the cell types Figure 4. Visualization of tissue infiltrating γδ T cells. Overview of a few antibodies included in the panel (left). Image of fresh, frozen human colon (centre). Examples of γδ T cell stains (right). Credit: Kilian Wistuba-Hamprecht, Manfred Claassen. Figure 5. Quantified neighborhoods of γδ T cells. Credit: Kilian Wistuba-Hamprecht, Manfred Claassen.4 initially identified. Therefore, if cell types were defined too broadly or too narrowly, no differential frequencies would be observed between healthy and diseased tissue. To overcome this challenge, the computer model was developed to identify possible novel cell types that are particularly depleted or enriched across tissue types independently – without accounting for the cell type annotations – instead characterizing them according to their molecular markers. This set of cells was then used as input for a shallow convolutional neural network – CellCnn – to classify the tissue type of origin of the input and enable the identification of potential novel cell types. When this model was applied to liver cancer data, the groups of cells enriched in the vicinity of MAIT cells in healthy, rim and tumor tissue could be observed, and the key enriched cell populations identified (Figure 7). Subsequent differential Figure 6. Cell-type annotation from multiplexed tissue images which were segmented to identify individual cells and their associated marker expression profiles. Credit: Kilian Wistuba-Hamprecht, Manfred Claassen. Figure 7. A & B) Identification of PD-L1+ M2 macrophages in MAIT neighborhoods. C) Mapping back the PD-L1+ M2 macrophage -MAIT interaction to the raw images to confirm the viability of the model. Credit: Ruf et al., 20235 expression analysis showed a strong enrichment of programmed cell death ligand-1 (PD-L1)+ M2 macrophages around MAIT cells. These M2 macrophages are known for their inhibitory effects on other immune cells, and the presence of PD-L1 on these macrophages suggests a mechanism along the PD-1/PD-L1 axis to confer inhibitory function. The mechanistic hypothesis of MAIT inhibition by PD-1/PDL1 mediation was further investigated in a mouse model. M2 macrophages were depleted, which resulted in a far stronger infiltration of the tumor and stronger effector functions of MAIT cells. This supports the hypothesis that M2 macrophages inhibit the ability of MAIT cells to fight liver cancer. Conclusion The MACSima Platform was used to establish and image an ultra-high-plex panel, including γδ T cell markers. This panel was then used to characterize γδ T cell microenvironments in CRC. In HCC tissue, the data produced by the MACSima Platform allowed analysis of the tumor-associated microenvironment of MAIT cells in HCC, and subsequently allowed the identification of a PD-L1+ M2 macrophage subset. The use of both machine learning analysis and MACSima Platform high-plex tissue imaging shows how collaborative experimental and computational methods can be used to establish a joint aim and investigate biological mechanisms. High-plex tissue imaging with the MACSima Platform is ideal to produce information rich data which can be used to form detailed mechanistic hypotheses to follow up in further experiments. Authors Dr. Kilian Wistuba-Hamprecht Dr. Wistuba-Hamprecht leads groups at the German Cancer Research Center (DKFZ) and University of Heidelberg. He led an independent research group at the University Hospital Tübingen with a focus on cancer immunology, monitoring the interplay between immune systems and tumors with phenotypic and functional investigations. Prof. Dr. Manfred Claassen Prof. Dr. Manfred Claassen is Professor for Clinical Bioinformatics at the University of Tübingen. His research focuses on learning explainable quantitative models in systems medicine from high dimensional single-cell omics data. References 1. Aktar N, Yueting C, Abbas M, et al. Understanding of immune escape mechanisms and advances in cancer immunotherapy. J Oncol. 2022:8901326. doi: 10.1155/2022/8901326 2. Girard P, Charles J, Cluzel C, et al. The features of circulating and tumorinfiltrating γδ T cells in melanoma patients display critical perturbations with prognostic impact on clinical outcome. Oncoimmunology. 2019;8(8):1601483. doi: 10.1080/2162402X.2019.1601483 3. Herold N, Bruhns M, Babaei S, et al. High-dimensional in situ proteomics imaging to assess γδ T cells in spatial biology. J Leukoc Biol. 2024;115(4):750–759. doi: 10.1093/jleuko/qiad167 4. Duan M, Goswami S, Shi J-Y, et al. Activated and exhausted MAIT cells foster disease progression and indicate poor outcome in hepatocellular carcinoma. Clin Cancer Res. 2019;25(11):3304–3316. doi: 10.1158/1078- 0432.CCR-18-3040